[WIP] Serialization of StagingPaths

[dwhswenson]

This is a subset of the work in #234, split off to facilitate review. This PR will merge into #258 (making it easier to see the additions specific to the PR, but meaning that this should be merged before that)

This handles the serialization of StagingPath objects. (detailed description will be coming in a future edit)

TODO:

• Add user story where the type of external storage is changed
• Add user story where the a new JSONCodec is added after the StagingSerialization is created.
• Docstrings
• Attach to the StorageManager

[pep8speaks]

Hello @dwhswenson! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

• In the file gufe/tests/storage/test_stagingserialization.py:

Line 68:1: E303 too many blank lines (3)

Comment last updated at 2023-12-22 16:58:26 UTC

[richardjgowers]

what's the need for the refresh_handler to be a method rather than just inlined into __init__?

or what's the harm in just adding to JSONCodec on the import of this module?

[dwhswenson]

what's the need for the refresh_handler to be a method rather than just inlined into init?

In principle, the codecs in JSON_HANDLER can change. For example, an external package can register additional handlers. Since So we at least want the ability to refresh. In fact, I was thinking about whether the encoder/decoder properties maybe should call refresh_handler.

or what's the harm in just adding to JSONCodec on the import of this module?

I'm not sure what you mean by "adding to JSONCodec." Could you clarify?

[richardjgowers]

Sorry I meant JSON_HANDLER. Other Codecs are defined and just appended onto the HANDLER as they are defined, so the HANDLER is "batteries included". Is there a reason this Codec can't have the same treatment?

[dwhswenson]

1. We don't know the storage paths at import time, so this is literally impossible.
2. We may have more than one set of shared/permanent (especially permanent) defined in a given process. Imagine that I'm working with data from multiple campaigns. Imagine furthermore that I've use the keep_shared option so that my shared stuff is also still accessible. This precludes any kind of serialization that relies on global state. This test is specifically to prevent us from making this mistake.